1. Field of The Invention
This invention relates to an apparatus and method for delivering involatile reagents, e.g., materials comprising Group II elements of the Periodic Table, in a form suitable for introduction into a deposition zone wherein films or layers are formed on a substrate by decomposition of the reagent.
2. Description of the Related Art
Recently many refractory materials have been identified as having unique materials properties. The recently discovered high temperature superconducting (HTSC) materials include YBa.sub.2 Cu.sub.3 O.sub.x, wherein x is from about 6 to 7.3, BiSrCaCuO, and TlBaCaCuO. Barium titanate, BaTiO.sub.3, has been identified as a ferroelectric and photonic material with unique and potentially very useful properties. Ba.sub.x Sr.sub.1-x Nb.sub.2 O.sub.6 is a photonic material whose index of refraction changes as a function of electric field and also as a function of the intensity of light upon it. Lead zirconate titanate, PbZr.sub.1-x Ti.sub.x O.sub.3, is a ferroelectric material whose properties are very interesting. The Group II metal fluorides, BaF.sub.2, CaF.sub.2, and SrF.sub.2, are materials that are useful for scintillation detecting and coating of optical fibers.
Many of the potential applications of these materials require their use in thin film, coating, or layer form. The films or layers may also be advantageously epitaxially related to the substrate upon which they are formed. Applications in which the refractory materials may need to be deposited in film or layer form include integrated circuits, switches, radiation detectors, thin film capacitors, holographic storage media, and various other microelectronic devices.
Uses that are currently envisioned for the copper oxide superconductors include high speed switches, bolometers, and high frequency communications components such as mixers. These applications would desirably use the HTSC material in the form of thin films in devices that can be made using well-known microelectronic fabrication technology.
Thin films comprising the Group II metal fluorides, BaF.sub.2, CaF.sub.2, and SrF.sub.2, are potentially very useful as buffer layers for interfacing between silicon substrates and HTSC or GaAs overlayers or between GaAs substrates and HTSC or silicon overlayers, and combinations of two or all of such metal fluorides may be employed in forming graded compositions in interlayers providing close lattice matching at the interfaces with the substrate and overlayer constituents of the composite. For example, a silicon substrate could be coated with an epitaxial layer of BaF.sub.2 /CaF.sub.2, SrF.sub.2 /CaF.sub.2, or SrF.sub.2 /CaF.sub.2 /BaF.sub.2, whose composition is tailored for a close lattice match to the silicon. If the ratio of the respective Group II metal species in the metal fluoride interlayers can be controlled precisely in the growth of the interlayer, the lattice constant could be graded to approach the lattice constant of GaAs. Thus, a gallium arsenide epitaxial layer could be grown over the metal fluoride interlayer, allowing the production of integrated GaAs devices on widely available, high quality silicon substrates. Another potential use of such type of metal fluoride interlayers would be as buffers between silicon substrates and polycrystalline HTSC films for applications such as non-equilibrium infrared detectors. Such an interlayer would permit the HTSC to be used in monolithic integrated circuits on silicon substrates.
BaTiO.sub.3 and Ba.sub.x Sr.sub.1-x Nb.sub.2 O.sub.6 in film or epitaxial layer form are useful in photonic applications such as optical switching, holographic memory storage, and sensors. In these applications, the BaTiO.sub.3 or Ba.sub.x Sr.sub.1-x Nb.sub.2 O.sub.6 film is the active element. The related ferroelectric material PbZr.sub.1-x Ti.sub.x O.sub.3 is potentially useful in infrared detectors and thin film capacitors well as filters and phase shifters.
Chemical vapor deposition (CVD) is a particularly attractive method for forming these layers because it is readily scaled up to production runs and because the electronics industry has a wide experience and an established equipment base in the use of CVD technology which can be applied to new CVD processes. In general, the control of key variables such as stoichiometry and film thickness, and the coating of a wide variety of substrate geometries is possible with CVD. Forming the thin films by CVD will permit the integration of these materials into existing device production technologies. CVD also permits the formation of layers of the refractory materials that are epitaxially related to substrates having close crystal structures.
CVD requires that the element source reagents must be sufficiently volatile to permit gas phase transport into the deposition reactor. The element source reagent must decompose in the reactor to deposit only the desired element at the desired growth temperatures. Premature gas phase reactions leading to particulate formation must not occur, nor should the source reagent decompose in the lines before reaching the reactor deposition chamber. When compounds, especially the HTSC materials, are desired to be deposited, obtaining optimal properties requires close control of stoichiometry which can be achieved if the reagent can be delivered into the reactor in a controllable fashion. Close control of stoichiometry would also be desired, for example, in the application described above involving graded Group II metal fluoride interlayers.
Many potentially highly useful refractory materials have in common that one or more of their components are elements, such as the Group II metals barium, calcium, and strontium, or early transition metals such as zirconium or hafnium, for which no volatile compounds well-suited for CVD are known. In many cases, the source reagents are solids which can be sublimed for gas-phase transport into the reactor. However, the sublimation temperature may be very close to the decomposition temperature, in which case the reagent may begin to decompose in the lines before reaching the reactor, and it will be very difficult to control the stoichiometry of the deposited films.
The formation of multicomponent refractory and electronic thin films such as high temperature superconductors [see K. Shinohara, F. Munakata, and M. Yamanaka, Jpn. J. Appl. Phys. 27, L1683 (1988); and K. Zhang, E. P. Boyd, B. S. Kwak, and A. Erbil, Appl. Phys. Lett. 55, 1258 (1989)] and ferroelectrics [see C. J. Brierley, C. Trundle, L. Considine, R. W. Whatmore, and F. W. Ainger, Ferroelectrics 91, 181 (1989); and F. W. Ainger, C. J. Brierley, M. D. Hudson, C. Trundle, and R. W. Whatmore, Ferroelectric Thin Films, E. R. Myers and A. I. Kingdon, Eds., Materials Research Society, Pittsburgh, Pa., 1990] by CVD or metalorganic CVD (MOCVD) has been only marginally successful. These materials have placed new demands on CVD technology, with the primary stumbling blocks being the inability to achieve exacting control of the stoichiometry of the multicomponent films and the absence of volatile source reagents. In many instances the source reagents are solids which decompose at temperatures near or slightly above the temperature at which they sublime.
Organogroup I and II complexes are particularly problematic. With the use of conventional bubblers, the bubbler is held at a temperature sufficiently high to sublime the reagent and consequently significant and somewhat variable decomposition of the source reagents occurs during a single growth run. This premature decomposition causes variations in the composition as a function of thickness in the as-deposited films and poor reproducibility in film stoichiometry between different growth runs.
Inexacting compositional control is particularly deleterious to high temperature superconducting thin films because the superconducting properties are extremely sensitive to the stoichiometry of the layer [see K. Wasa, H. Adachi, Y. Ichikawa, K. Setsune, and K. Hirochi, Review Solid State Sci. 2, 453 (1988); and R. F. Bunshsh, and C. V. Deshpandey, Research and Development 65 (1989)]. Two approaches involving the use of nonconventional hardware have been tried to overcome this problem. The first method eliminates the bubblers and uses a reactor tube which contains concentric tubes, each containing a boat filled with a single source reagent. A temperature gradient is applied along the tube to vaporize each material at the required temperature [see M. Ihara, T. Kimura, H. Yamawaki, and K. Ikeda, IEEE Trans. Magnetics 25, 2471 (1989); and H. Yamane, H. Kurosawa, T. Hirai, K. Watanabe, H. Iwasaki, N. Kobayashi, and Y. Muto, J. Crystal Growth 98, 860 (1989]. There are several drawbacks to this method: (1) as with standard bubblers, significant decomposition occurs during a given run because the reagents are held at high temperatures for the duration of the run; (2) temperature control is not as good as with standard bubblers, thus giving rise to wide variations in source reagent vapor pressure and consequently to wide variations in the stoichiometry of the as-deposited films, and (3) the boats need to be charged before each run, a step which is not consistent with a high volume commercial process.
The second method uses two bubblers in series. The first bubbler contains a volatile chelating ligand which presumably acts to stablize and/or to lower the melting point of the source reagent which is contained in the second (downstream) bubbler, [see P. H. Dickinson, T. H. Geballe, A. Sanjurjo, D. Hildenbrand, G. Craig, M. Zisk, J. Collman, S. A. Banning, and R. E. Sievers, J. Appl. Phys, 66 444 (1989)]. Stabilities on the order of a few hours which are sufficient for a single run have been realized with this method. However, a fresh charge of source reagent is needed before each run. In addition, some enhancement in the vapor pressure of the source reagent occurs. Unfortunately, the amount of enhancement is not reproducible, which again causes variations in the stoichiometry of the as-deposited films.
Standard CVD processes, such as the deposition of tungsten metallization layers from tungsten hexafluoride and silane [see R. S. Rosler, J. Mendonca, M. J. Rice, Jr., J. Vac, Sci. Technol., B6, 1721 (1988)], which use gaseous or liquid source reagents, are not plagued by source reagent decomposition. Furthermore, deposition conditions (i.e., substrate temperature, reactor pressure, and partial pressures of the source reagents) can typically be found where the growth of the desired phase is kinetically or thermodynamically favored over individual impurity phases. For example, the MOCVD of single crystal GaAs thin films is carried out with an arsine+trimethylgallium mixture in which there is typically a 30 to 60-fold excess of the group V source reagent Under appropriately selected conditions, the excess arsine either does not react or the extra arsenic evaporates before it can be incorporated into the rowing film [see T. F. Kuech, Mat. Sci. Reports, S. S. Lau and F. W. Saris, Eds 2, 1 (1987)].
By contrast, during CVD processes for refractory materials, the vapor pressures of the binary oxides, nitrides, and carbides are often lower than that of the desired multicomponent phase at the deposition temperature. Thus any excess source reagent leads to the deposition of a binary oxide or carbide which is then incorporated as an impurity phase in the growing film.
In summary, the techniques heretofore employed for formation of refractory thin films from relatively involatile reagents have not permitted efficient delivery of the reagents into the reactor and close control of reagent ratios and hence film stoichiometry.
Other objects and advantages of the present invention will be more fully apparent from the ensuing disclosure and appended claims.